A few years ago I started an open source project called mondrian. I had been building some business intelligence applications, and had taken a liking to OLAP. I was hoping to see OLAP applications popping up here and there, on websites, in desktop applications like Quicken, and it just wasn't happening.
I soon figured out why. OLAP was cool, but it was complicated and the software was expensive. These two facts are not unrelated. If you're a software vendor who has just developed a cool new technology, you're going to want to charge a lot of money for it. But your customer, who is paying a lot of money for it, wants to see a lot more in the box than a CD and some packing peanuts. So, the software tends to get expensive, to match the customer's expectations.
Now, don't get me wrong. Business Intelligence is a complicated process. It involves getting a business person to talk to a computer person, which is something neither of them enjoys very much. In fact there are probably several business people, and systems administrators, and lawyers, and consultants to get them all to talk to each other, and lots and lots of data.
But does the software have to be complicated too? Not really, I thought.
So, mainly to prove a point, I wrote the mondrian OLAP server.
And things worked out fine. Today a lot of people use mondrian, and mondrian does a lot of things. In the early days, it was a combination of engineering pragmatism, dumb luck, and dogged persistence.
The engineering pragmatism was that since I was one guy, working code at the weekend, this thing had better be really simple. Java and XML. Could Java possibly be fast enough? Well, it had better be, I thought. Someone's gone to the trouble of writing a DBMS already, and it's great at storing data and crunching numbers, so let's make the database do as much of the work as possible.
The dumb luck was that on the other side of the world, almost at exactly the same time I brought out the first prototype, Andreas Voss was thinking of writing an OLAP client which would fit perfectly with the server I was writing. That client was jpivot, of course. An OLAP server isn't much use without a client to do ad hoc queries.
The dogged persistence was not so much me, writing the code, as the initial set of users, using mondrian and jpivot for heavens knows what, but using it anyway, and fixing bugs, and just getting involved.
By the way, I'm still not sure exactly what mondrian users do with mondrian. It's one of the peculiar side-effects of open source. Because no money changes hands, because there's no purchase order or legal agreement, the relationship with the users is fleeting. You bump into some of them in bars at conferences, and of course you hear some war stories, but typically you see a few French table names in a bug report, and think "Cool. Someone's using Mondrian to read electricity meters. Who'd have thought?"
But the main thing is, mondrian just gets better and better. And my job as the lead developer is to stop it from getting too complicated, while it gets better. I'll have more to say about that in the next few posts.